Avoid blocking in arc_reclaim_thread()

As described in the comment above arc_reclaim_thread() it's critical
that the reclaim thread be careful about blocking.  Just like it must
never wait on a hash lock, it must never wait on a task which can in
turn wait on the CV in arc_get_data_buf().  This will deadlock, see
issue #3822 for full backtraces showing the problem.

To resolve this issue arc_kmem_reap_now() has been updated to use the
asynchronous arc prune function.  This means that arc_prune_async()
may now be called while there are still outstanding arc_prune_tasks.
However, this isn't a problem because arc_prune_async() already
keeps a reference count preventing multiple outstanding tasks per
registered consumer.  Functionally, this behavior is the same as
the counterpart illumos function dnlc_reduce_cache().

Signed-off-by: Brian Behlendorf <[email protected]>
Signed-off-by: Tim Chase <[email protected]>
Issue #3808
Issue #3834
Issue #3822

1 files changed, 4 insertions, 11 deletions

diff --git a/module/zfs/arc.c b/module/zfs/arc.c
index 7cd4e76f2..b759e6483 100644
--- a/module/zfs/arc.c
+++ b/module/zfs/arc.c
@@ -2685,8 +2685,8 @@ arc_flush_state(arc_state_t *state, uint64_t spa, arc_buf_contents_t type,
 }
 
 /*
- * Helper function for arc_prune() it is responsible for safely handling
- * the execution of a registered arc_prune_func_t.
+ * Helper function for arc_prune_async() it is responsible for safely
+ * handling the execution of a registered arc_prune_func_t.
  */
 static void
 arc_prune_task(void *ptr)
@@ -2711,7 +2711,7 @@ arc_prune_task(void *ptr)
  * honor the arc_meta_limit and reclaim otherwise pinned ARC buffers.  This
  * is analogous to dnlc_reduce_cache() but more generic.
  *
- * This operation is performed asyncronously so it may be safely called
+ * This operation is performed asynchronously so it may be safely called
  * in the context of the arc_reclaim_thread().  A reference is taken here
  * for each registered arc_prune_t and the arc_prune_task() is responsible
  * for releasing it once the registered arc_prune_func_t has completed.
@@ -2736,13 +2736,6 @@ arc_prune_async(int64_t adjust)
 	mutex_exit(&arc_prune_mtx);
 }
 
-static void
-arc_prune(int64_t adjust)
-{
-	arc_prune_async(adjust);
-	taskq_wait_outstanding(arc_prune_taskq, 0);
-}
-
 /*
  * Evict the specified number of bytes from the state specified,
  * restricting eviction to the spa and type given. This function
@@ -3376,7 +3369,7 @@ arc_kmem_reap_now(void)
 		 * We are exceeding our meta-data cache limit.
 		 * Prune some entries to release holds on meta-data.
 		 */
-		arc_prune(zfs_arc_meta_prune);
+		arc_prune_async(zfs_arc_meta_prune);
 	}
 
 	for (i = 0; i < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; i++) {